BASIC ELECTRONIC PROPERTIES OF a-Si:H/c-Si HETEROSTRUCTURE SOLAR CELLS M. Schmidt1, L. Korte1, K. Kliefoth1, A. Schoepke1, R. Stangl1, A. Laades1, E. Conrad1 , K. Brendel1, W. Fuhs1 M. Scherff2, W. Fahrner2 1
Hahn-Meitner-Institut Berlin, Abteilung Silizium-Photovoltaik, Kekuléstr. 5, D-12489 Berlin Tel: +49/30/8062-1352, Fax: +49/30/8062-1333, e-mail:
[email protected] 2
University of Hagen, Department of Electrical Engineering Haldener Str. 182, D-58084 Hagen, Tel.: ++49-(0)2331/987-4012, email:
[email protected]
ABSTRACT: Amorphous/crystalline silicon heterostructures with 2-10 nm thin a-Si:H (i,n,p) layers have been prepared by PECVD deposition and analyzed in respect of their basic properties. The gap state densities of the thin a-Si:H layers have been determined by UV-excited photoelectron spectroscopy. The concentration of deep states depends on the doping level and lies in the range between 1018-1020 cm-3. The Fermi level, resulting from UPS analysis, shifts up to 1.47 eV above the valence band at 10000ppm PH3 addition to SiH4. At a-Si:H layer thicknesses below 3nm the onset of the photoelectron contribution from the c-Si valence band gave a value of the valence band offset of 450 ±50meV. The c-Si band bending induced by differently doped a-Si:H layers was determined by surface photovoltage measurements, SPV. The interface state density of the a-Si:H/c-Si interface reaches values down to 2 1011 cm-2eV-1 at midgap. These results were obtained from field dependent SPV investigations which indicate an excellent passivation of the interface by the a-Si:H network. Finally, using the optimized preparation parameters, we prepared a ITO/a-Si:H(n)/c-Si(p)/BSF hetero solar cell with an efficiency of 17 %. Keywords: 1:c-Si - 2:a-Si - 3:Heterojunction
Heterostructures have some inherent advantages compared to conventional homo pn junctions because they allow to combine the useful properties of two solids. The heterojunction between amorphous hydrogenated silicon and crystalline silicon, a-Si:H/c-Si, represents such a system which can be prepared at temperatures below 300°C, allowing the application on thin low cost silicon wafers or silicon films on glass. The common challenge of the heterostructures results from the fact that the first monolayers of the heterotransition between both solids determine their properties [1] like the interface state density distribution Dit(E), the band offsets between the corresponding valence bands ∆EV and the conduction bands ∆EC and the band bending qϕSO in the crystalline silicon c-Si. These quantities are sketched in Fig. 1. c-Si:H(p):
16
12
EF qϕS0 N(E) ∆ EV Dit(E)
Figure 1: Band scheme of a TCO/a-Si:H/c-Si heterojunction solar cell; interface state density Dit(E), gap state density N(E), band offsets ∆EC and ∆EV, and band bending qϕSO. The aim of this paper is to show how these basic properties influence the solar cell performance, how the quantities can be measured and what their actual values are. This provides the basis for realistic numerical simulations of such heterojunctions by the AFORS-HET
-2
-1
12
-2
-1
Dit=10 cm eV
10 8
4
∆ EC
10
Dit=10 cm eV
14
LD= 400µm
nptype
6
a-Si:H(n)+ TCO metal
programme [2] and for preparation of optimised heterosolar cells. The potential of this cell type on n-type silicon was demonstrated by the a-Si:H(p)/a-Si:H(i)/cSi(n) pn-heterojunction solar cell from Sanyo where the cell efficiency η amounts to 17,3% for flat wafers and 20,7 % for structured c-Si wafers [3,4]. For the inverse type the Si:H(n)/c-Si(p) np-heterojunction solar cell the reached values amount to η=16.2% on flat c-Si wafers [5], which has been improved up to 17%, as shown in this paper.
η [%]
1. INTRODUCTION
0
pntype
100 200 300 400 500 600 ∆EC,V [meV]
Figure 2: Numerically calculated dependence of the aSi:H/c-Si hetero solar cell efficiency η on the bandoffset for the minority charge carriers ∆EC,V for two values of interface state density. The loss in efficiency due to interface recombination is highlighted by the shaded areas for the two cell types. Fig. 2 makes evident that the most important parameters of this type of solar cells is the a-Si:H/c-Si heterojunction with its interface state density Dit(E) and the bandoffset ∆EV,C of the minority charge carriers. Their transport across the heterojunction becomes limited at offset values above 450 meV. This corresponds to the ∆EV value of hole transfer at the a-Si:H(p)/c-Si(n) junction. On the other hand, the band offset fixes the Fermi level position directly at the interface and thus determines the interface recombination rate at a given Dit(E). The minority band offset for the a-Si:H(n)/c-Si(p) junction goes down to about ∆EC ≈100meV [6] which results in an increased
interface recombination at a given Dit(E). This emphasizes that the determination of ∆EC,V and Dit(E) are central tasks for physical insights as well as for device development.
2. SAMPLE PREPARATION All samples were prepared on c-Si substrates (2” or 3”FZ, 1-75 Ωcm) with (111) surface orientation. After cleaning the wafers by the standard RCA process and a HF-dip (1% HF, 60 sec) the substrates were inserted into the CVD deposition chamber within 5 minutes. The passivation of the substrate surface (H-termination) by the HF-dip is known to be stable against oxidation for about 30 minutes. The a-Si:H layers were deposited by plasma enhanced chemical vapor deposition (PECVD) at 13.56 MHz and 15-55 mW/cm2 power density. Process gases were semiconductor grade silane (SiH4) phosphine (PH3) or diboran (B2H6) and pure hydrogen. Gas flow, Hdilution and the sample temperature during deposition (170°C...250°C) were varied for layer and interface optimization. Immediately after deposition, the samples were transferred to the photoelectron spectroscopy analysis chamber without breaking the vacuum or were stored in dry nitrogen for SPV measurements. For the preparation of n/p-type heterojunction solar cells we used flat, 350 µm thick, (111) oriented, monocrystalline FZ silicon wafers (0.5 – 2 Ωcm) with a diffused back suface field (BSF). After the wafer cleaning, as desribed above, the a-Si:H (n) layer was deposited. After that, an indium tin oxide (ITO) layer with a thickness of 80 nm, an integral transmission of 86 % and a specific resistance of 3x10-4Ωcm was deposited by dc-sputtering from a compound target. Finally, the front and back contacts were prepared by thermal evaporation. 30 nm Cr and 3 µm Ag layers structured with photolithography are used on the ITO and 2 µm Al (full area) is deposited on the backside. Cells with an area of 1 cm2 are prepared by wet chemical etching the ITO in order to separate the cells. 3. EXPERIMENTAL METHODS 3.1 Photoelectron Spectroscopy (PES) The energetic distribution of the occupied gap states and upper valence band states NOC(E) and the Fermi level position were determined by UV excited (hν=3-7 eV) photoemission. The photoemission yield is determined by the distributions of the occupied and unoccupied states, NOC(E) and NU(E), and the optical transition matrix element [7]. Assuming a constant density of states (DOS) of the unoccupied final states NU(E) [8] and assuming that the optical matrix element remains nearly constant up to hν = 3.5 eV and decreases above this value as shown in [9], the measured yield Y(E), corrected by the optical matrix element, directly traces NOC(E). For excitation energies below 10eV the inelastic mean free path, λimfp, of the photoelectrons, depending on the interaction with phonons, electrons and plasmons, strongly raises up to about 5nm at 6eV because energy losses by plasmon generation processes are impossible [10]. The detection limit of the electron emission depth can be assumed to be about 3λimfp. Therefore, excitation in the energy range between the work function limit at 3.8 eV and about 7 eV allows the measurement of the
DOS of ultrathin a-Si:H layers with a thickness of 5-7 nm . The photoelectrons were excited by strong monochromatic UV-light (3 – 7.5 eV) generated by passing the light of a Xe high pressure lamp through a double-grating monochromator. In this low excitation energy range, the number of absorbed photons was measured using calibrated silicon diodes for the detection of the incoming and reflected light beams. This procedure allows to measure the absolute photoelectron quantum yield Y(E) at each photon energy in two different modes. In the first mode, corresponding to standard UPS, the distribution of the kinetic energy of the photoelectrons was detected during excitation with a fixed photon energy. In the other mode, the energy analyzer was operated at a fixed energy (final state energy) while varying the photon energy (3-7 eV), a technique known as Constant Final State Yield Spectroscopy (CFSYS) [8]. The energy resolution of the analyzer was better than 100 meV. 3.1 Surface photovoltage (SPV) The surface photovoltage method allows to determine both the band bending in the c-Si absorber and the interface state density distribution. For that purpose the change in band bending caused by optically generated charge carriers [11] is monitored and the initial band bending is changed by applying an external voltage. The decay with time of excess charge carriers after their generation in c-Si by a light pulse (λ=934nm, 150ns pulse length) is measured as the voltage decay of an artificial MIS structure. The artificial MIS structure consists of the glas/TCO/mica/a-Si:H/c-Si/Al sample where the top (metal) electrode is a TCO covered glas plate followed by mica as an insulator. The solution of Poisson's equation for the space charge region at the (free) sample surface (c-Si) yields the space charge density of the band bending region QSC = const.⋅F(ϕ0,λ,δp), where F(ϕ0,λ,δp)-doping function, ϕ0initial band bending, λ=p/ni - doping factor, δp - excess charge [12]. To determine the initial band bending from the doping function presumes that no recharging during the light pulse takes place, i.e. F0(ϕ0,λ,δp=0) = F(ϕ,λ,δp≠0). This means that during the light exposure (pulse length ~150 ns), no states at the a-Si:H/c-Si interface or close to it change their charge state. In this case the photovoltage is given by the band bending change before and at the end of illumination, corrected by the Dember voltage UD, which results from the different mobilities of the generated electrons and holes ; Uph= (ϕ-ϕ0)- UD. The determination of the interface state density distribution needs the correctly measured band bending ϕS and their relation to the applied external gate voltage Ug. The method for determining Dit(ϕS) is given in detail in [13].
4.
RESULTS AND DISCUSSION
First the UPS and CFSYS modes were used to determine the position of the Fermi level with respect to the valence band edge EF - EV, the slope of the valence band tail states (Urbach energy EO) and the variation of the distribution of dangling bond (db) states with doping. The Fermi level position results directly from the zero point of the binding energy EB. From the slope of the straight line in a plot of √Y(E) vs. E the onset of the
parabolic density of states distribution at the valence band edge is determined which defines the value of EV. At this energy the density of states amounts to NOV ≈ 2 ×1021 cm-3 eV-1 [8]. This value was used here to define a quantitative scale for the DOS distribution. The Urbach energy EOV of the exponentially distributed valence band tail states was obtained from the slope of a log-linear plot of NOC(E) vs. E. The obtained value of EOV reflects the disorder broadening of the valence band with doping as shown from curves (1), (2) and (3) in Fig. 3. The EOV amounts to 91 meV, 136 meV and 120meV, respectively. These are characteristic values for doped a-Si:H layers [14]. Additionally, it can be seen that the deep defect state density decreases for slightly doped compared to the n-doped layers by nearly one order of magnitude. The deep defect density of curve (3) is not measurable because only occupied states can be detected by PES according to N(E)OC= N(E) F(E), where F(E) is the Fermi-Dirac distribution function. 1E22 a-Si:H (p) a-Si:H (i) a-Si:H (n)
-1
N(E)OC [cm eV ]
1E21 -3
1E20 1E19 (3)
1E18
(1)
(2)
1E17 1E16
band edges of a-Si:H and the solid lines the c-Si band edges, respectively. The basic characteristics of the gap state distribution are nearly unchanged when the layer thickness is reduced to values down to 2.8 nm (Fig. 4). This was confirmed by measurements on different series of samples deposited with varying thickness on c-Si (111), c-Si (100) and also on Ag-coated wafers. It is surprising that the density of deep defects changes only very little even in the thinnest sample. However, there is clearly a change of the slope of the valence band tail. EOV increases from 73 meV for the thick layers to 101 meV for the ultrathin layer (2.8 nm) (Fig.4). Based on the comparison with NOC(E) spectra of 2nm thin a-Si:H on an Ag substrate we believe that this results from contributions of the crystalline silicon substrate to the photoemission. A measurable contribution of photoelectrons from the c-Si substrate can be expected only for a-Si:H layer thickness d
